Put review time in the pilot funding request
A promising AI demo is one input to a funding decision. The work around it belongs in the request too.
Who reviews the output? How much paid human handling remains? What corrections follow that first review? Which recurring charges depend on usage? Which future setup and operating costs can still be avoided if the pilot stops?
Record those inputs separately. Do not count the same minutes as handling, review and rework. Keep unattended model runtime separate from paid human effort. A precise-looking total remains conditional on the inputs behind it.
Capacity value and cash effects also answer different questions. Time released for other work is not automatically cash saved. A pilot can show an economic benefit under stated assumptions while still requiring cash to operate. Make that distinction visible rather than combining both into a single confident savings claim.
Alongside the base case, record what would change the choice: a higher review burden, different usage, missing evidence, an operating dependency or a cost beyond the approved limit. Name the decision owner and the response when a stop condition is reached. A spreadsheet result is information for that owner, not spending authority.
The AI Pilot Funding Decision Pack supplies a two-page decision note, full-cost model, evidence/assumption workbook and pilot/stop-condition documents, with fictional worked examples. You provide the evidence, choose assumptions and own the funding decision. No professional recommendation, founder analysis or guaranteed return is included.